Professional Documents
Culture Documents
SSRN Id2567732
SSRN Id2567732
SSRN Id2567732
Affiliations:
*Christian von Haldenwang
(corresponding author)
German Development Institute (Deutsches Institut für Entwicklungspolitik, DIE)
Tulpenfeld 6
53113 Bonn
Germany
Christian.vonHaldenwang@die-gdi.de
Tel: +49 (0) 228-94927 282
** Maksym Ivanyna
Joint Vienna Institute
mivanyna@jvi.org
Abstract: Conventional wisdom has it that natural resource dependence is associated with
increased vulnerability to external shocks emanating from sudden supply and demand
changes in the global economy. This paper explores to what extent government revenue in
low- and middle-income countries is affected by different kinds of shocks and whether these
effects are different for resource-rich as compared to non-resource-rich countries. Based on
data from 176 countries between 1980-2010, we measure the elasticity of tax revenue in
resource-rich countries with respect to two kinds of shocks: exchange rate pressure and
terms of trade shocks. We find that government revenue in resource-rich countries is more
volatile, but not necessarily more vulnerable, in particular with regard to terms of trade
shocks. Poorer resource-rich countries are more vulnerable than their higher-income
counterparts. Vulnerability in resource-rich countries has significantly decreased in the
2000s as compared to previous decades. Introducing institutional variables such as political
regime type or bureaucratic quality we find that the general institutional characteristics of a
country may not always reflect the quality of its management of natural resources. Still,
results provide some support to those arguments that stress the relevance of good
governance in the context of the so-called “resource curse”.
Acknowledgements: This paper is partly based on a broader study on the vulnerability and
resilience factors of tax revenues in developing countries (von Haldenwang et al., 2013). Co-
authors of the study are Oliver Morrissey, Ingo Bordon and Armin von Schiller. Financial
support of the European Commission is gratefully acknowledged.
Electronic copy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2567732
http://ssrn.com/abstract=2567732
Vulnerability of tax revenue in resource-rich countries
1. Introduction
Conventional wisdom has it that countries with a high dependence on natural resource
exports are particularly vulnerable to exogenous shocks. Their economies usually depend on
a small range of export products with oscillating world market prices, making public revenue
from these sources highly volatile. In addition, abundance of natural resources often leads to
the crowding-out of investments in other sectors (the famous ‘Dutch disease’), resulting in
small tax bases.
But is this conventional wisdom corroborated by empirical findings? To be sure, a broad
body of literature links abundance of natural resources to volatility of growth and revenue
(Crivelli and Gupta, 2014; IMF, 2013). But volatility does not equal vulnerability. In this
paper, ‘volatility’ is taken as a measure of revenue instability, based on deviations around an
observed (linear or exponential) trend. ‘Vulnerability’ refers to the elasticity of tax revenue
with respect to different kinds of exogenous shocks. Concerning the latter, empirical
research is rather scarce, especially with regard to developing countries.
In theory, resource-rich countries that are able to deal with volatile revenue sources through
sound macroeconomic and fiscal management could be even less affected by external
shocks than other countries. Or, put the other way round: High revenue vulnerability of
resource-rich countries could also depend on other, including domestic, factors, such as for
instance wide-spread rent-seeking, corruption, limited state capacity (Thies, 2010) or
economic mismanagement (Raddatz, 2007). To give an example, government revenue from
mining in Zambia has oscillated between 5 and 18 per cent between 2002 and 2011, while at
the same time the average government revenue take compared to the contribution of
mining to GDP was less than a fifth of the ratios obtained by international benchmark
countries such as Chile or Botswana. Zambia’s poor performance has been linked to political
instability and a critical lack of state capacity in the management of the mining sector
(Lundstøl et al., 2013).
Hence, the first puzzle this paper sets out to explore is whether natural resource
dependence is associated with increased vulnerability to external shocks. We are particularly
interested in the fate of poorer countries, as it can be supposed that they will find it more
difficult to implement the sound policies mentioned above. The second puzzle analysed in
the paper refers to the influence of governance factors on vulnerability. Following the
resource curse literature we expect political regime type and state capacity to impact on the
way governments deal with resource wealth (Andersen and Ross, 2011; Collier and Hoeffler,
2009; Davis, 2013; Ehrhart and Guerineau, 2013; Haber and Menaldo, 2011; Liou and
Musgrave, 2012; McGuirk, 2013; Ross, 2012a; Schaffer and Ziyadov, 2012; Wright et al.,
2014).
Electronic copy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2567732
http://ssrn.com/abstract=2567732
The paper explores the effect of two kinds of shocks on tax revenue in a total of 176
countries – the exchange rate pressure index, a proxy for export demand and foreign capital
flows shocks, and terms-of-trade shocks. It compares resource-rich (RR) and non-resource-
rich (non-RR) countries1 regarding their revenue volatility and exposure to shocks. Further, it
analyses whether countries behave differently according to income levels, political regime
type and government effectiveness.
We find evidence that revenues of RR-countries are indeed more volatile than those of non-
RR countries. They also face more volatile shocks – in particular with regard to terms of
trade. Looking at the impact of shocks, however, a differentiated picture emerges. RR
countries are not necessarily more vulnerable to external shocks. While exchange rate
pressure affects RR countries more than non-RR countries, the effect is driven above all by
poorer (low- and lower-middle-income) countries. With regard to terms-of-trade shocks, RR
and non-RR countries are equally vulnerable and in both groups, revenues of poorer
countries are more affected by this kind of shock. We also observe that RR as well as non-RR
countries reduce the vulnerability of their revenue to terms-of-trade shocks in the 2000s,
compared to the previous two decades. In the case of RR countries, this could be the result
of a general improvement of resource management.
Finally, the introduction of governance-related indicators produces mixed results. For
instance, democracies appear to fare better facing exchange rate pressure shocks than non-
democracies in RR countries, whereas results are inconclusive for non-RR countries. In
contrast, terms-of-trade shocks seem to affect countries independently of their governance
structure. The results suggest that the low quality of governance in general does not
necessarily translate into poor governance of the resource sector, which is a major source of
public revenue in resource-rich countries.
The paper proceeds as follows: The next section develops the argument based on a
discussion of the literature on shocks, revenue volatility and natural resources. Section 3
introduces the method and data used. Section 4 presents the findings of the econometric
analysis. Section 5 concludes.
2. Literature review
The impact of natural resources on the volatility of economic growth is a well-established
fact (for developing countries, see for instance Arezki et al., 2012). Many observations point
to a strong relationship between volatility and vulnerability. For instance, in a recent study
on managing volatility, the International Monetary Fund observes: “research suggests that
external shocks contribute to large output losses and protracted growth slowdowns in LICs”
(IMF, 2013: 7). The IMF paper continues: “A number of LICs face fragilities defined by their
weak institutions, ongoing or recent conflict, and high poverty levels, which put them in a
weak position to cope with the effects of shocks and to mediate their social impact. [...] Such
1
To identify RR countries we rely on the categorisation introduced by the IMF (2012). For a list of RR
countries included in the analysis please refer to the Appendix.
2
It should be noted, however, that the sample consists mainly of high- and upper-middle-income countries.
3
The micro-theoretical mechanisms underlying this relationship are complex and difficult to measure,
however. Introducing additional variables, various authors find that the negative effect of resource-based
revenue on democracy or democratization weakens or disappears completely. For instance, Herb (2005)
observes that other factors (region, Muslim share of the population and income levels) have stronger
effects on political regime type than the resource rent. Morrison (2009) and Liou and Musgrave (2012)
arrive at the conclusion that resource rents stabilise political regimes of any kind. In a study on 15 sub-
Saharan countries, McGuirk (2013) shows that rents from natural resources lead to lower levels of tax
enforcement, thus reducing the demand for democratic accountability. Haber and Menaldo (2011) even
report the opposite effect of resource rents, but their approach has attracted criticism pointing to
conceptual and methodological flaws (see Andersen and Ross, 2011).
3. Methodological approach
The paper explores the impact of two different kinds of shocks on tax revenue in a broad
range of countries. We take data from 176 countries, covering the period 1980-2010. First,
we build measures of fiscal capacity and revenue volatility4 for each country to characterize
tax resilience features of the sample. These measures are regressed on a set of control
factors in order to account for structural features of the economies. In identifying the
control variables we follow the tax effort literature (for instance, see Bird, 1976; Fenochietto
and Pessino, 2013; Gupta, 2007; Tanzi, 1992; Teera and Hudson, 2004). Most of these
studies use sector shares (mainly agriculture and manufacturing) and a composite measure
of trade openness, such as the share of trade volume (imports plus exports) on GDP.
However, sector shares are less suitable for panel analysis with annual data or short period
averages, as changes are usually rather slow and follow a clear trend. Further, we assume
4
The measure of volatility used in this paper is described in Appendix 2.
where i identifies the country, t is the year, E is the exchange rate in local currency units per
USD, RES – size of reserves, and are country-specific weights:
. Here is the standard deviation of in country i in 1980-2012,
is the same for .
The logic behind the index is that in response to an adverse balance-of-payment shock a
country could employ different strategies: the government could devaluate the currency,
5
We also check for possible non-linear effects when the magnitude of shocks is particularly large. For each
shock X, we define a dummy variable “X, large”, which is equal to 1 if a shock is greater than the 90th
percentile of the respective income group distribution. See IMF (2013) for a similar approach. However, we
find no evidence for non-linear effects of shocks on revenue. See von Haldenwang et al. (2013) for more
details.
where i is the country index, and t is the year index; rev is total revenue without grants (as
percentage to GDP); w is external shock - ToT or ER_pressure X is the vector of our controls
– agriculture exports, mineral exports, fuel exports, manufactured exports and imports,
agricultural value-added and GDP per capita. is a random error. Our interest is β. For any
variable a, ä denotes its time-demeaned value:
6
The weights in (2) are country-specific and chosen so that the more volatile series gets smaller weight. To
reduce the impact of outliers, the ER pressure index is transformed as follows:
7
In addition to these two shocks, we check for GDP decline as a proxy for a general output shock and for
intensity of natural disasters, a measure based on people killed and affected by natural disaster in every
year t and every country j (Fomby et al., 2009). With regard to GDP decline, all coefficients are statistically
insignificant and very close to zero. These results indicate that on average tax systems are neutral, i.e. the
elasticity of revenue with respect to output is close to 1. With regard to natural disasters, we find that RR
countries face less severe natural disasters shocks than non-RR countries, and their revenues are not
significantly affected by them. The reason could be that extractive industries – the main sources of revenue
in RR countries – are usually less affected by natural catastrophes than other types of economic activity,
such as for instance agriculture. For a more in-depth analysis of these relationships, see von Haldenwang et
al. (2013).
In analyzing the effect of shocks on revenue we account for the level of welfare using the
World Bank country income groups and dividing the sample in high- and upper-middle-
income countries (the “richer” countries) on the one hand, and lower-middle- and low-
income countries (the “poorer” countries) on the other hand.
As an additional focus we explore the behaviour of resource-rich vs. non-resource-rich
countries in different time periods – in particular before and after the year 2000. This follows
the idea that many resource-rich countries were able to improve their management of the
extractive sector in recent years, partly due to the long commodity supercycle and the
lessons learnt from previous decades.
We then divide the sample according to governance characteristics, in order to tease out
possible effects of the “political resource curse”. First we split countries according to their
political regime characteristics, based on their Polity IV score: On a scale ranging from 10 to -
10, a country is considered a democracy if its Polity IV score is higher than 5, and a non-
democracy if otherwise (see Marshall et al., 2010). As a second measure we use the
bureaucratic quality index from the ICRG data set.
4. Findings
Our findings show that the revenue structure of resource-rich countries, though more
volatile than that of non-resource-rich countries due to a higher dependence on non-tax
revenue, could be less vulnerable to external shocks. At least, there is no robust evidence
pointing to volatility of revenue from natural resources being directly connected to
increased vulnerability vis-à-vis external shocks.
First, as indicated by higher mean values, government revenue is on average more volatile in
RR countries. This is true for the whole sample of 1980-2010, as well as for the 1980s, 90s,
and 2000s separately. The difference in the volatility index between RR and non-RR
countries is largest in the 2000s (14.3 vs. 5.22), and smallest in the 90s (9.47 vs. 5.71).
Second, RR countries seem to be more heterogeneous, as indicated by the higher standard
deviation of the volatility index, as well as the spread between the 10th and 90th percentiles
of the index distribution. Again, the heterogeneity among RR countries is largest in the
8
This may be related to greater dependency of these countries to natural resources in the 2000s.
9
Note that the terms-of-trade shocks are in logarithms.
When accounting for natural resource endowments, however, the results fail to provide
clear evidence for a better performance of democracies. Owing perhaps to small sample
size, evidence is indicative at best and could be driven by outliers – at least in the group of
16 RR democracies. For RR countries, findings point to a lower sensitivity of democratic
regimes with regard to ER pressure shocks. In contrast, the point estimates of the effects of
ToT shocks on government revenue are higher for democratic regimes.
Looking at non-RR countries, there is no evidence of differentiated patterns for democratic
or non-democratic regimes. It appears noteworthy, however, that the difference between
the political regimes is not very large, and about the same in RR and non-RR countries. On
the one hand, in non-RR countries tax effort and trust in government are likely to matter
more for revenue collection and the fight against tax evasion, which makes political regime a
more relevant indicator for fiscal performance in these countries. On the other hand,
transparency and accountability of revenue collection from extractive activities could matter
in RR countries as well. As shown by other studies (see Garcia and von Haldenwang, 2015),
stable autocratic regimes may be as efficient in collecting taxes as democratic governments.
The second subgroup division that we consider is by time period – 1980-2000 vs. 2001-2010.
The results are presented in Table 4. Again, results mostly fail to reach statistical significance,
but the picture that emerges is that RR countries managed their government revenue better
in the 2000s than in the earlier period – the effects of exchange rate pressure and terms-of-
trade shocks are essentially zero in 2000s. This may be the result of a general improvement
We also divide countries by the quality of their public sector (based on the bureaucratic
quality index of ICRG). Results are presented in Table 5. As it appears, bureaucratic quality
seems to be more relevant for the resilience to ER pressure shocks than to ToT shocks. Table
5 shows that in the RR group ToT shocks affect tax revenue in high bureaucratic quality
countries more than in countries with low bureaucratic quality. However, the result may
once again be driven by outliers, as there are only ten RR countries with high bureaucratic
quality in our sample. For non-RR countries low bureaucratic quality generally makes
countries less resilient to external shocks, judging from the magnitudes of the coefficients,
but only the group with low bureaucratic quality produces statistically significant results. All
in all the results seem to reflect the ambiguous nature of empirical findings regarding the
effect of natural resource wealth on administrative capacity, reported in Section 2 of this
paper.
5. Concluding remarks
Countries rich in natural resources tend to suffer from higher degrees of revenue volatility,
but this is not always a consequence of their exposure to exogenous shocks. Of the two
kinds of shocks we have analyzed in this paper – exchange rate pressure and terms-of-trade
shocks – the former seems to affect both RR and non-RR countries, albeit differently. Among
the RR countries, those belonging to the lower income group are more vulnerable than the
higher-income countries. At the same time, RR countries (especially the poorer ones) appear
to be less affected by ToT shocks.
This finding is even more interesting as the shocks themselves are more volatile for RR
countries than for non-RR countries. The standard deviation of both shocks – and of ToT
shocks in particular – is higher for RR countries. In this regard conventional wisdom appears
to be corroborated by the facts: Countries with a high dependency on natural resources face
higher oscillations of world market prices, on average. Somewhat surprisingly, however, we
2. Measuring volatility
where y is the variable whose level of instability we want to find, t is time, and et is the
residual at time t. To obtain the index we use the formula:
n
yt yˆt
2
100 t1
(I)
y n3
where y is the arithmetic mean of y, y t is the observed value of y in year t and yˆ t is the
estimated value of y, from (2), in year t. The square root term in (I) yields the standard
deviation of residuals from a quadratic time trend, as the mean of the residuals is necessarily
zero. This is divided by the arithmetic mean of y to normalise the index, enabling cross-
country comparisons to be made. The variable t2 is included in the time trend to pick up
possible non-linearities. The index is to be interpreted as the typical deviation of the variable
from a quadratic time trend over the period. As such it records average volatility over this
period.
Table 1: Revenue volatility and size of shocks in RR and non-RR countries, by decade
Total Government Exchange Rate
Terms of Trade
Revenue Volatility Pressure
Non-RR RR Non-RR RR Non-RR RR
1980-2010
Mean 7.91 17.03 -0.16 -0.17 -0.02 -0.13
Standard deviation 6.20 18.46 2.21 2.46 0.23 0.38
10th percentile 2.49 5.70 -2.90 -3.19 -0.25 -0.63
90th percentile 13.47 46.15 2.78 3.13 0.20 0.23
Observations 117 51 3082 1280 2298 1190
1981-1990
Mean 7.51 12.00 0.33 0.58 -0.08 -0.25
Standard deviation 10.23 13.96 2.37 2.57 0.36 0.42
10th percentile 1.90 2.47 -2.93 -3 -0.54 -0.81
90th percentile 13.28 27.52 3.16 3.59 0.29 0.26
Observations 58 16 863 353 508 290
1991-2000
Mean 5.63 9.56 0.21 0.24 -0.04 0.05
Standard deviation 4.47 6.14 2.18 2.45 0.15 0.31
10th percentile 1.53 3.26 -2.68 -3.09 -0.2 -0.23
90th percentile 11.43 18.14 2.82 3.16 0.12 0.45
Observations 101 35 1048 431 572 325
2001-2010
Mean 4.83 12.78 -0.84 -1.06 0.02 -0.17
Standard deviation 3.44 17.14 1.92 2.08 0.16 0.34
10th percentile 1.49 2.81 -2.98 -3.3 -0.13 -0.59
90th percentile 8.92 49.61 2.05 2.14 0.2 0.1
Observations 102 48 1171 496 1170 550
Note: RR = resource-rich
ER pressure shock -0.134*** -0.159** -0.115** -0.160* 0.174 -0.192* -0.0984** -0.175** -0.0335
(0.0419) (0.0651) (0.0539) (0.0883) (0.172) (0.107) (0.0474) (0.0692) (0.0625)
ToT shock -4.404*** -2.469*** -4.625*** -4.221*** -2.286 -3.388** -4.311*** -1.445 -4.786***
(0.481) (0.771) (0.626) (1.062) (2.177) (1.467) (0.658) (1.201) (0.767)
agr_exp_togdp -0.0312 0.0135 -0.00543 0.236** 0.0267 0.149 -0.0855*** -0.125*** -0.0972**
(0.0283) (0.0417) (0.0413) (0.103) (0.489) (0.110) (0.0290) (0.0420) (0.0437)
min_exp_togdp -0.0489 0.102 -0.142** 0.0235 0.217 -0.194** 0.0390 0.266* -0.221
(0.0542) (0.0924) (0.0667) (0.0790) (0.159) (0.0961) (0.0989) (0.139) (0.140)
manuf_exp_togdp -0.0623*** -0.0220 -0.122*** 0.144* 0.426*** -0.112 -0.111*** -0.158*** -0.117***
(0.0171) (0.0232) (0.0285) (0.0862) (0.161) (0.113) (0.0177) (0.0259) (0.0281)
imports_togdp 0.0470*** -0.0445* 0.142*** 0.0509 -0.0573 0.0948** 0.0841*** 0.0705*** 0.161***
(0.0149) (0.0230) (0.0203) (0.0396) (0.0967) (0.0439) (0.0168) (0.0265) (0.0224)
fuel_exp_togdp 0.168*** 0.132*** 0.335*** 0.317*** 0.367*** 0.348*** -0.0945* -0.274*** 0.460***
(0.0352) (0.0486) (0.0524) (0.0545) (0.0790) (0.0770) (0.0546) (0.0730) (0.0967)
agri_va_gdp -0.0486** -0.104 -0.0451* -0.163** -0.0585 -0.0327 -0.0227 0.210** -0.0469*
(0.0236) (0.101) (0.0239) (0.0676) (0.484) (0.0698) (0.0241) (0.106) (0.0241)
ln_gdp 0.230 -3.498*** 3.065*** -1.776 -8.518*** 9.823*** 0.474 0.701 0.814
(0.718) (1.195) (0.959) (1.585) (2.197) (2.608) (0.826) (1.433) (1.065)
constant 23.25*** 64.80*** -3.754 31.44*** 98.23*** -49.89*** 22.21*** 25.60* 12.47
(5.809) (11.10) (6.923) (12.05) (21.22) (17.80) (6.789) (13.30) (7.766)
R-squared 0.219 0.202 0.326 0.414 0.646 0.479 0.178 0.206 0.315
observations 1680 769 911 428 142 286 1252 627 625
Number of id 139 68 71 40 14 26 99 54 45
Note: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1; HIC = high-income countries; UMIC = upper-middle-income countries; LMIC =
lower-middle-income countries; LIC = low-income countries; RR = resource-rich